US 12,248,814 B2
Method and apparatus for accelerating GNN pre-processing
Myoungsoo Jung, Daejeon (KR); Seungkwan Kang, Daejeon (KR); Donghyun Gouk, Daejeon (KR); Miryeong Kwon, Daejeon (KR); Hyunkyu Choi, Daejeon (KR); and Junhyeok Jang, Daejeon (KR)
Assigned to KOREA ADVANCED INSTITUTE OF SCIENCE AND TECHNOLOGY, Daejeon (KR)
Filed by KOREA ADVANCED INSTITUTE OF SCIENCE AND TECHNOLOGY, Daejeon (KR)
Filed on Aug. 22, 2023, as Appl. No. 18/453,702.
Claims priority of application No. 10-2023-0030199 (KR), filed on Mar. 7, 2023.
Prior Publication US 2024/0303122 A1, Sep. 12, 2024
Int. Cl. G06F 7/76 (2006.01); G06F 7/36 (2006.01); G06F 9/50 (2006.01)
CPC G06F 9/5027 (2013.01) [G06F 7/36 (2013.01)] 14 Claims
OG exemplary drawing
 
1. A method of accelerating graph neural network (GNN) pre-processing, the method comprising:
performing, by a set-partitioning accelerator, radix sorting based on a vertex identification (VID) of an original graph in a coordinate list (COO) format to generate a COO array of a preset length;
merging, by a merger, the COO array of the preset length to generate one sorted COO array;
converting, by a converter, the one sorted COO array into a compressed sparse row (CSR) format to generate a graph in a CSR format;
performing, by a set-partitioning accelerator, uniform random sampling, to generate a sub-graph with a reduced degree of the graph in the CSR format; and
generating, by an embedding table generation unit, an embedding table corresponding to the sub-graph,
wherein when the merger is provided as a plurality of mergers, pipelining is performed among the plurality of mergers.